Spaces:
Runtime error
Runtime error
Upload folder using huggingface_hub
Browse files- a.wav +0 -0
- emotion-recognition-wav2vec2-IEMOCAP/hyperparams.yaml +2 -2
- facebook/wav2vecChinese/README.md +61 -0
- facebook/wav2vecChinese/config.json +115 -0
- facebook/wav2vecChinese/gitattributes.txt +27 -0
- facebook/wav2vecChinese/hyperparams.yaml +59 -0
- facebook/wav2vecChinese/preprocessor_config.json +9 -0
- facebook/wav2vecChinese/pytorch_model.bin +3 -0
- hyperparams.yaml +65 -0
- out.wav +0 -0
- paraformer_logs/log/info.log +51 -0
- pretrained_path/custom_interface.py +221 -0
- pretrained_path/label_encoder.ckpt +3 -0
- pretrained_path/label_encoder.txt +4 -0
- pretrained_path/model.ckpt +3 -0
- pretrained_path/wav2vec2.ckpt +3 -0
- speech.py +1 -1
a.wav
CHANGED
Binary files a/a.wav and b/a.wav differ
|
|
emotion-recognition-wav2vec2-IEMOCAP/hyperparams.yaml
CHANGED
@@ -9,10 +9,10 @@ HPARAMS_NEEDED: ["encoder_dim", "out_n_neurons", "label_encoder", "softmax"]
|
|
9 |
MODULES_NEEDED: ["wav2vec2", "avg_pool", "output_mlp", "LSTM"]
|
10 |
|
11 |
# Feature parameters
|
12 |
-
wav2vec2_hub:
|
13 |
|
14 |
# Pretrain folder (HuggingFace)
|
15 |
-
pretrained_path:
|
16 |
|
17 |
# parameters
|
18 |
encoder_dim: 1024
|
|
|
9 |
MODULES_NEEDED: ["wav2vec2", "avg_pool", "output_mlp", "LSTM"]
|
10 |
|
11 |
# Feature parameters
|
12 |
+
wav2vec2_hub: facebook\wav2vecChinese
|
13 |
|
14 |
# Pretrain folder (HuggingFace)
|
15 |
+
pretrained_path: pretrained_path
|
16 |
|
17 |
# parameters
|
18 |
encoder_dim: 1024
|
facebook/wav2vecChinese/README.md
ADDED
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
---
|
4 |
+
Pretrained on 10k hours WenetSpeech L subset. More details in [TencentGameMate/chinese_speech_pretrain](https://github.com/TencentGameMate/chinese_speech_pretrain)
|
5 |
+
|
6 |
+
This model does not have a tokenizer as it was pretrained on audio alone.
|
7 |
+
In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data.
|
8 |
+
|
9 |
+
python package:
|
10 |
+
transformers==4.16.2
|
11 |
+
|
12 |
+
```python
|
13 |
+
|
14 |
+
|
15 |
+
import torch
|
16 |
+
import torch.nn.functional as F
|
17 |
+
import soundfile as sf
|
18 |
+
from fairseq import checkpoint_utils
|
19 |
+
|
20 |
+
from transformers import (
|
21 |
+
Wav2Vec2FeatureExtractor,
|
22 |
+
Wav2Vec2ForPreTraining,
|
23 |
+
Wav2Vec2Model,
|
24 |
+
)
|
25 |
+
from transformers.models.wav2vec2.modeling_wav2vec2 import _compute_mask_indices
|
26 |
+
|
27 |
+
model_path=""
|
28 |
+
wav_path=""
|
29 |
+
mask_prob=0.0
|
30 |
+
mask_length=10
|
31 |
+
|
32 |
+
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained(model_path)
|
33 |
+
model = Wav2Vec2Model.from_pretrained(model_path)
|
34 |
+
|
35 |
+
# for pretrain: Wav2Vec2ForPreTraining
|
36 |
+
# model = Wav2Vec2ForPreTraining.from_pretrained(model_path)
|
37 |
+
|
38 |
+
model = model.to(device)
|
39 |
+
model = model.half()
|
40 |
+
model.eval()
|
41 |
+
|
42 |
+
wav, sr = sf.read(wav_path)
|
43 |
+
input_values = feature_extractor(wav, return_tensors="pt").input_values
|
44 |
+
input_values = input_values.half()
|
45 |
+
input_values = input_values.to(device)
|
46 |
+
|
47 |
+
# for Wav2Vec2ForPreTraining
|
48 |
+
# batch_size, raw_sequence_length = input_values.shape
|
49 |
+
# sequence_length = model._get_feat_extract_output_lengths(raw_sequence_length)
|
50 |
+
# mask_time_indices = _compute_mask_indices((batch_size, sequence_length), mask_prob=0.0, mask_length=2)
|
51 |
+
# mask_time_indices = torch.tensor(mask_time_indices, device=input_values.device, dtype=torch.long)
|
52 |
+
|
53 |
+
with torch.no_grad():
|
54 |
+
outputs = model(input_values)
|
55 |
+
last_hidden_state = outputs.last_hidden_state
|
56 |
+
|
57 |
+
# for Wav2Vec2ForPreTraining
|
58 |
+
# outputs = model(input_values, mask_time_indices=mask_time_indices, output_hidden_states=True)
|
59 |
+
# last_hidden_state = outputs.hidden_states[-1]
|
60 |
+
|
61 |
+
```
|
facebook/wav2vecChinese/config.json
ADDED
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"activation_dropout": 0.0,
|
3 |
+
"adapter_kernel_size": 3,
|
4 |
+
"adapter_stride": 2,
|
5 |
+
"add_adapter": false,
|
6 |
+
"apply_spec_augment": true,
|
7 |
+
"architectures": [
|
8 |
+
"Wav2Vec2ForPreTraining"
|
9 |
+
],
|
10 |
+
"attention_dropout": 0.1,
|
11 |
+
"bos_token_id": 1,
|
12 |
+
"classifier_proj_size": 256,
|
13 |
+
"codevector_dim": 768,
|
14 |
+
"contrastive_logits_temperature": 0.1,
|
15 |
+
"conv_bias": true,
|
16 |
+
"conv_dim": [
|
17 |
+
512,
|
18 |
+
512,
|
19 |
+
512,
|
20 |
+
512,
|
21 |
+
512,
|
22 |
+
512,
|
23 |
+
512
|
24 |
+
],
|
25 |
+
"conv_kernel": [
|
26 |
+
10,
|
27 |
+
3,
|
28 |
+
3,
|
29 |
+
3,
|
30 |
+
3,
|
31 |
+
2,
|
32 |
+
2
|
33 |
+
],
|
34 |
+
"conv_stride": [
|
35 |
+
5,
|
36 |
+
2,
|
37 |
+
2,
|
38 |
+
2,
|
39 |
+
2,
|
40 |
+
2,
|
41 |
+
2
|
42 |
+
],
|
43 |
+
"ctc_loss_reduction": "sum",
|
44 |
+
"ctc_zero_infinity": false,
|
45 |
+
"diversity_loss_weight": 0.1,
|
46 |
+
"do_stable_layer_norm": true,
|
47 |
+
"eos_token_id": 2,
|
48 |
+
"feat_extract_activation": "gelu",
|
49 |
+
"feat_extract_dropout": 0.0,
|
50 |
+
"feat_extract_norm": "layer",
|
51 |
+
"feat_proj_dropout": 0.1,
|
52 |
+
"feat_quantizer_dropout": 0.0,
|
53 |
+
"final_dropout": 0.0,
|
54 |
+
"gradient_checkpointing": false,
|
55 |
+
"hidden_act": "gelu",
|
56 |
+
"hidden_dropout": 0.1,
|
57 |
+
"hidden_size": 1024,
|
58 |
+
"initializer_range": 0.02,
|
59 |
+
"intermediate_size": 4096,
|
60 |
+
"layer_norm_eps": 1e-05,
|
61 |
+
"layerdrop": 0.1,
|
62 |
+
"mask_channel_length": 10,
|
63 |
+
"mask_channel_min_space": 1,
|
64 |
+
"mask_channel_other": 0.0,
|
65 |
+
"mask_channel_prob": 0.0,
|
66 |
+
"mask_channel_selection": "static",
|
67 |
+
"mask_feature_length": 10,
|
68 |
+
"mask_feature_min_masks": 0,
|
69 |
+
"mask_feature_prob": 0.0,
|
70 |
+
"mask_time_length": 10,
|
71 |
+
"mask_time_min_masks": 2,
|
72 |
+
"mask_time_min_space": 1,
|
73 |
+
"mask_time_other": 0.0,
|
74 |
+
"mask_time_prob": 0.075,
|
75 |
+
"mask_time_selection": "static",
|
76 |
+
"model_type": "wav2vec2",
|
77 |
+
"num_adapter_layers": 3,
|
78 |
+
"num_attention_heads": 16,
|
79 |
+
"num_codevector_groups": 2,
|
80 |
+
"num_codevectors_per_group": 320,
|
81 |
+
"num_conv_pos_embedding_groups": 16,
|
82 |
+
"num_conv_pos_embeddings": 128,
|
83 |
+
"num_feat_extract_layers": 7,
|
84 |
+
"num_hidden_layers": 24,
|
85 |
+
"num_negatives": 100,
|
86 |
+
"output_hidden_size": 1024,
|
87 |
+
"pad_token_id": 0,
|
88 |
+
"proj_codevector_dim": 768,
|
89 |
+
"tdnn_dilation": [
|
90 |
+
1,
|
91 |
+
2,
|
92 |
+
3,
|
93 |
+
1,
|
94 |
+
1
|
95 |
+
],
|
96 |
+
"tdnn_dim": [
|
97 |
+
512,
|
98 |
+
512,
|
99 |
+
512,
|
100 |
+
512,
|
101 |
+
1500
|
102 |
+
],
|
103 |
+
"tdnn_kernel": [
|
104 |
+
5,
|
105 |
+
3,
|
106 |
+
3,
|
107 |
+
1,
|
108 |
+
1
|
109 |
+
],
|
110 |
+
"torch_dtype": "float32",
|
111 |
+
"transformers_version": "4.16.2",
|
112 |
+
"use_weighted_layer_sum": false,
|
113 |
+
"vocab_size": 32,
|
114 |
+
"xvector_output_dim": 512
|
115 |
+
}
|
facebook/wav2vecChinese/gitattributes.txt
ADDED
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
5 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
6 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
7 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
8 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
9 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
10 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
11 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
12 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
13 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
14 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
15 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
16 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
17 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
18 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
19 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
20 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
21 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
22 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
23 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
24 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
25 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
26 |
+
*.zstandard filter=lfs diff=lfs merge=lfs -text
|
27 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
facebook/wav2vecChinese/hyperparams.yaml
ADDED
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# ############################################################################
|
2 |
+
# Model: WAV2VEC base for Emotion Recognition
|
3 |
+
# ############################################################################
|
4 |
+
|
5 |
+
|
6 |
+
# Hparams NEEDED
|
7 |
+
HPARAMS_NEEDED: ["encoder_dim", "out_n_neurons", "label_encoder", "softmax"]
|
8 |
+
# Modules Needed
|
9 |
+
MODULES_NEEDED: ["wav2vec2", "avg_pool", "output_mlp"]
|
10 |
+
|
11 |
+
# Feature parameters
|
12 |
+
wav2vec2_hub: wav2vecChinese
|
13 |
+
|
14 |
+
# Pretrain folder (HuggingFace)
|
15 |
+
pretrained_path: emotion-recognition-wav2vec2-IEMOCAP
|
16 |
+
|
17 |
+
# parameters
|
18 |
+
encoder_dim: 768
|
19 |
+
out_n_neurons: 4
|
20 |
+
|
21 |
+
wav2vec2: !new:speechbrain.lobes.models.huggingface_wav2vec.HuggingFaceWav2Vec2
|
22 |
+
source: D:/pycharm2020/code/yuyin_ChineseWav2vec/pretrained_models/facebook/wav2vec2-base
|
23 |
+
output_norm: True
|
24 |
+
freeze: True
|
25 |
+
save_path: wav2vec2_checkpoints
|
26 |
+
|
27 |
+
avg_pool: !new:speechbrain.nnet.pooling.StatisticsPooling
|
28 |
+
return_std: False
|
29 |
+
|
30 |
+
output_mlp: !new:speechbrain.nnet.linear.Linear
|
31 |
+
input_size: !ref <encoder_dim>
|
32 |
+
n_neurons: !ref <out_n_neurons>
|
33 |
+
bias: False
|
34 |
+
|
35 |
+
model: !new:torch.nn.ModuleList
|
36 |
+
- [!ref <output_mlp>]
|
37 |
+
|
38 |
+
modules:
|
39 |
+
wav2vec2: !ref <wav2vec2>
|
40 |
+
output_mlp: !ref <output_mlp>
|
41 |
+
avg_pool: !ref <avg_pool>
|
42 |
+
|
43 |
+
softmax: !new:speechbrain.nnet.activations.Softmax
|
44 |
+
|
45 |
+
|
46 |
+
label_encoder: !new:speechbrain.dataio.encoder.CategoricalEncoder
|
47 |
+
|
48 |
+
|
49 |
+
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
|
50 |
+
loadables:
|
51 |
+
wav2vec2: !ref <wav2vec2>
|
52 |
+
model: !ref <model>
|
53 |
+
label_encoder: !ref <label_encoder>
|
54 |
+
paths:
|
55 |
+
wav2vec2: !ref <pretrained_path>/wav2vec2.ckpt
|
56 |
+
model: !ref <pretrained_path>/model.ckpt
|
57 |
+
label_encoder: !ref <pretrained_path>/label_encoder.txt
|
58 |
+
|
59 |
+
|
facebook/wav2vecChinese/preprocessor_config.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"do_normalize": true,
|
3 |
+
"feature_extractor_type": "Wav2Vec2FeatureExtractor",
|
4 |
+
"feature_size": 1,
|
5 |
+
"padding_side": "right",
|
6 |
+
"padding_value": 0,
|
7 |
+
"return_attention_mask": true,
|
8 |
+
"sampling_rate": 16000
|
9 |
+
}
|
facebook/wav2vecChinese/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c8a5554a79c3bbbe76f2e43d3d4b4369c8c2abd5515e623192e0381d7e5e7b3f
|
3 |
+
size 1269726951
|
hyperparams.yaml
ADDED
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# ############################################################################
|
2 |
+
# Model: WAV2VEC base for Emotion Recognition
|
3 |
+
# ############################################################################
|
4 |
+
|
5 |
+
|
6 |
+
# Hparams NEEDED
|
7 |
+
HPARAMS_NEEDED: ["encoder_dim", "out_n_neurons", "label_encoder", "softmax"]
|
8 |
+
# Modules Needed
|
9 |
+
MODULES_NEEDED: ["wav2vec2", "avg_pool", "output_mlp", "LSTM"]
|
10 |
+
|
11 |
+
# Feature parameters
|
12 |
+
wav2vec2_hub: facebook\wav2vecChinese
|
13 |
+
|
14 |
+
# Pretrain folder (HuggingFace)
|
15 |
+
pretrained_path: emotion-recognition-wav2vec2-IEMOCAP
|
16 |
+
|
17 |
+
# parameters
|
18 |
+
encoder_dim: 1024
|
19 |
+
out_n_neurons: 2
|
20 |
+
|
21 |
+
wav2vec2: !new:speechbrain.lobes.models.huggingface_wav2vec.HuggingFaceWav2Vec2
|
22 |
+
source: !ref <wav2vec2_hub>
|
23 |
+
output_norm: True
|
24 |
+
freeze: True
|
25 |
+
save_path: wav2vec2_checkpoints
|
26 |
+
|
27 |
+
LSTM: !new:speechbrain.nnet.RNN.LSTM
|
28 |
+
input_size: !ref <encoder_dim>
|
29 |
+
hidden_size: !ref <encoder_dim>
|
30 |
+
|
31 |
+
avg_pool: !new:speechbrain.nnet.pooling.StatisticsPooling
|
32 |
+
return_std: False
|
33 |
+
|
34 |
+
output_mlp: !new:speechbrain.nnet.linear.Linear
|
35 |
+
input_size: !ref <encoder_dim>
|
36 |
+
n_neurons: !ref <out_n_neurons>
|
37 |
+
bias: False
|
38 |
+
|
39 |
+
model: !new:torch.nn.ModuleList
|
40 |
+
- [!ref <output_mlp>]
|
41 |
+
|
42 |
+
modules:
|
43 |
+
wav2vec2: !ref <wav2vec2>
|
44 |
+
output_mlp: !ref <output_mlp>
|
45 |
+
avg_pool: !ref <avg_pool>
|
46 |
+
LSTM: !ref <LSTM>
|
47 |
+
|
48 |
+
softmax: !new:speechbrain.nnet.activations.Softmax
|
49 |
+
log_softmax: !new:speechbrain.nnet.activations.Softmax
|
50 |
+
apply_log: True
|
51 |
+
|
52 |
+
label_encoder: !new:speechbrain.dataio.encoder.CategoricalEncoder
|
53 |
+
|
54 |
+
|
55 |
+
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
|
56 |
+
loadables:
|
57 |
+
wav2vec2: !ref <wav2vec2>
|
58 |
+
model: !ref <model>
|
59 |
+
label_encoder: !ref <label_encoder>
|
60 |
+
paths:
|
61 |
+
wav2vec2: !ref <pretrained_path>/wav2vec2.ckpt
|
62 |
+
model: !ref <pretrained_path>/model.ckpt
|
63 |
+
label_encoder: !ref <pretrained_path>/label_encoder.txt
|
64 |
+
|
65 |
+
|
out.wav
CHANGED
Binary files a/out.wav and b/out.wav differ
|
|
paraformer_logs/log/info.log
CHANGED
@@ -391,3 +391,54 @@
|
|
391 |
[INFO] 2024-02-19 11:37:47,247 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create PuncOrtInferRuntimeSession instance finished
|
392 |
[INFO] 2024-02-19 11:37:47,248 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create CT_Transformer instance finished
|
393 |
[INFO] 2024-02-19 11:37:47,249 [D:\DD\paraformer\runtime\python\cttPunctuator.py:26] __init__: Offline punctuator instance initialized.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
391 |
[INFO] 2024-02-19 11:37:47,247 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create PuncOrtInferRuntimeSession instance finished
|
392 |
[INFO] 2024-02-19 11:37:47,248 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create CT_Transformer instance finished
|
393 |
[INFO] 2024-02-19 11:37:47,249 [D:\DD\paraformer\runtime\python\cttPunctuator.py:26] __init__: Offline punctuator instance initialized.
|
394 |
+
[INFO] 2024-02-19 11:41:21,395 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating Campplus instance
|
395 |
+
[INFO] 2024-02-19 11:41:22,000 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create Campplus instance finished
|
396 |
+
[INFO] 2024-02-19 11:41:25,167 [D:\DD\paraformer\runtime\python\paraformerInfer.py:51] __init__: Load onnx model dir at D:\DD\paraformer\onnx\asr_offline
|
397 |
+
[INFO] 2024-02-19 11:41:25,168 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating ParaformerOfflineModel instance
|
398 |
+
[INFO] 2024-02-19 11:41:25,175 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating AsrOfflineOrtInferRuntimeSession instance
|
399 |
+
[INFO] 2024-02-19 11:41:28,326 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create AsrOfflineOrtInferRuntimeSession instance finished
|
400 |
+
[INFO] 2024-02-19 11:41:28,326 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create ParaformerOfflineModel instance finished
|
401 |
+
[INFO] 2024-02-19 11:41:28,333 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating VadOrtInferRuntimeSession instance
|
402 |
+
[INFO] 2024-02-19 11:41:28,366 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create VadOrtInferRuntimeSession instance finished
|
403 |
+
[INFO] 2024-02-19 11:41:28,367 [D:\DD\paraformer\runtime\python\cttPunctuator.py:24] __init__: Initializing punctuator instance with offline mode.
|
404 |
+
[INFO] 2024-02-19 11:41:28,368 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating CT_Transformer instance
|
405 |
+
[INFO] 2024-02-19 11:41:28,369 [D:\DD\paraformer\runtime\python\model\punc\punctuator.py:52] __init__: Loading config file D:\DD\paraformer\onnx\punc\config.pkl
|
406 |
+
[INFO] 2024-02-19 11:41:28,438 [D:\DD\paraformer\runtime\python\model\punc\punctuator.py:56] __init__: Loading config file D:\DD\paraformer\onnx\punc\config.pkl finished, takes 0.06800723075866699 s
|
407 |
+
[INFO] 2024-02-19 11:41:28,513 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating PuncOrtInferRuntimeSession instance
|
408 |
+
[INFO] 2024-02-19 11:41:32,325 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create PuncOrtInferRuntimeSession instance finished
|
409 |
+
[INFO] 2024-02-19 11:41:32,326 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create CT_Transformer instance finished
|
410 |
+
[INFO] 2024-02-19 11:41:32,327 [D:\DD\paraformer\runtime\python\cttPunctuator.py:26] __init__: Offline punctuator instance initialized.
|
411 |
+
[INFO] 2024-02-19 12:20:59,536 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating Campplus instance
|
412 |
+
[INFO] 2024-02-19 12:21:00,086 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create Campplus instance finished
|
413 |
+
[INFO] 2024-02-19 12:21:03,528 [D:\DD\paraformer\runtime\python\paraformerInfer.py:51] __init__: Load onnx model dir at D:\DD\paraformer\onnx\asr_offline
|
414 |
+
[INFO] 2024-02-19 12:21:03,529 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating ParaformerOfflineModel instance
|
415 |
+
[INFO] 2024-02-19 12:21:03,534 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating AsrOfflineOrtInferRuntimeSession instance
|
416 |
+
[INFO] 2024-02-19 12:21:06,482 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create AsrOfflineOrtInferRuntimeSession instance finished
|
417 |
+
[INFO] 2024-02-19 12:21:06,482 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create ParaformerOfflineModel instance finished
|
418 |
+
[INFO] 2024-02-19 12:21:06,490 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating VadOrtInferRuntimeSession instance
|
419 |
+
[INFO] 2024-02-19 12:21:06,540 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create VadOrtInferRuntimeSession instance finished
|
420 |
+
[INFO] 2024-02-19 12:21:06,542 [D:\DD\paraformer\runtime\python\cttPunctuator.py:24] __init__: Initializing punctuator instance with offline mode.
|
421 |
+
[INFO] 2024-02-19 12:21:06,542 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating CT_Transformer instance
|
422 |
+
[INFO] 2024-02-19 12:21:06,544 [D:\DD\paraformer\runtime\python\model\punc\punctuator.py:52] __init__: Loading config file D:\DD\paraformer\onnx\punc\config.pkl
|
423 |
+
[INFO] 2024-02-19 12:21:06,626 [D:\DD\paraformer\runtime\python\model\punc\punctuator.py:56] __init__: Loading config file D:\DD\paraformer\onnx\punc\config.pkl finished, takes 0.08150649070739746 s
|
424 |
+
[INFO] 2024-02-19 12:21:06,711 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating PuncOrtInferRuntimeSession instance
|
425 |
+
[INFO] 2024-02-19 12:21:10,434 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create PuncOrtInferRuntimeSession instance finished
|
426 |
+
[INFO] 2024-02-19 12:21:10,435 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create CT_Transformer instance finished
|
427 |
+
[INFO] 2024-02-19 12:21:10,436 [D:\DD\paraformer\runtime\python\cttPunctuator.py:26] __init__: Offline punctuator instance initialized.
|
428 |
+
[INFO] 2024-02-19 12:24:23,118 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating Campplus instance
|
429 |
+
[INFO] 2024-02-19 12:24:23,663 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create Campplus instance finished
|
430 |
+
[INFO] 2024-02-19 12:24:27,366 [D:\DD\paraformer\runtime\python\paraformerInfer.py:51] __init__: Load onnx model dir at D:\DD\paraformer\onnx\asr_offline
|
431 |
+
[INFO] 2024-02-19 12:24:27,367 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating ParaformerOfflineModel instance
|
432 |
+
[INFO] 2024-02-19 12:24:27,371 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating AsrOfflineOrtInferRuntimeSession instance
|
433 |
+
[INFO] 2024-02-19 12:24:30,101 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create AsrOfflineOrtInferRuntimeSession instance finished
|
434 |
+
[INFO] 2024-02-19 12:24:30,102 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create ParaformerOfflineModel instance finished
|
435 |
+
[INFO] 2024-02-19 12:24:30,109 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating VadOrtInferRuntimeSession instance
|
436 |
+
[INFO] 2024-02-19 12:24:30,158 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create VadOrtInferRuntimeSession instance finished
|
437 |
+
[INFO] 2024-02-19 12:24:30,159 [D:\DD\paraformer\runtime\python\cttPunctuator.py:24] __init__: Initializing punctuator instance with offline mode.
|
438 |
+
[INFO] 2024-02-19 12:24:30,160 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating CT_Transformer instance
|
439 |
+
[INFO] 2024-02-19 12:24:30,162 [D:\DD\paraformer\runtime\python\model\punc\punctuator.py:52] __init__: Loading config file D:\DD\paraformer\onnx\punc\config.pkl
|
440 |
+
[INFO] 2024-02-19 12:24:30,244 [D:\DD\paraformer\runtime\python\model\punc\punctuator.py:56] __init__: Loading config file D:\DD\paraformer\onnx\punc\config.pkl finished, takes 0.08200550079345703 s
|
441 |
+
[INFO] 2024-02-19 12:24:30,327 [D:\DD\paraformer\runtime\python\utils\singleton.py:26] get_instance: creating PuncOrtInferRuntimeSession instance
|
442 |
+
[INFO] 2024-02-19 12:24:33,888 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create PuncOrtInferRuntimeSession instance finished
|
443 |
+
[INFO] 2024-02-19 12:24:33,889 [D:\DD\paraformer\runtime\python\utils\singleton.py:29] get_instance: create CT_Transformer instance finished
|
444 |
+
[INFO] 2024-02-19 12:24:33,889 [D:\DD\paraformer\runtime\python\cttPunctuator.py:26] __init__: Offline punctuator instance initialized.
|
pretrained_path/custom_interface.py
ADDED
@@ -0,0 +1,221 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from speechbrain.pretrained import Pretrained
|
3 |
+
|
4 |
+
|
5 |
+
class CustomEncoderWav2vec2Classifier(Pretrained):
|
6 |
+
"""A ready-to-use class for utterance-level classification (e.g, speaker-id,
|
7 |
+
language-id, emotion recognition, keyword spotting, etc).
|
8 |
+
|
9 |
+
The class assumes that an self-supervised encoder like wav2vec2/hubert and a classifier model
|
10 |
+
are defined in the yaml file. If you want to
|
11 |
+
convert the predicted index into a corresponding text label, please
|
12 |
+
provide the path of the label_encoder in a variable called 'lab_encoder_file'
|
13 |
+
within the yaml.
|
14 |
+
|
15 |
+
The class can be used either to run only the encoder (encode_batch()) to
|
16 |
+
extract embeddings or to run a classification step (classify_batch()).
|
17 |
+
```
|
18 |
+
|
19 |
+
Example
|
20 |
+
-------
|
21 |
+
>>> import torchaudio
|
22 |
+
>>> from speechbrain.pretrained import EncoderClassifier
|
23 |
+
>>> # Model is downloaded from the speechbrain HuggingFace repo
|
24 |
+
>>> tmpdir = getfixture("tmpdir")
|
25 |
+
>>> classifier = EncoderClassifier.from_hparams(
|
26 |
+
... source="speechbrain/spkrec-ecapa-voxceleb",
|
27 |
+
... savedir=tmpdir,
|
28 |
+
... )
|
29 |
+
|
30 |
+
>>> # Compute embeddings
|
31 |
+
>>> signal, fs = torchaudio.load("samples/audio_samples/example1.wav")
|
32 |
+
>>> embeddings = classifier.encode_batch(signal)
|
33 |
+
|
34 |
+
>>> # Classification
|
35 |
+
>>> prediction = classifier .classify_batch(signal)
|
36 |
+
"""
|
37 |
+
|
38 |
+
def __init__(self, *args, **kwargs):
|
39 |
+
super().__init__(*args, **kwargs)
|
40 |
+
|
41 |
+
def encode_batch(self, wavs, wav_lens=None, normalize=False):
|
42 |
+
"""Encodes the input audio into a single vector embeddin g.
|
43 |
+
|
44 |
+
The waveforms should already be in the model's desired format.
|
45 |
+
You can call:
|
46 |
+
``normalized = <this>.normalizer(signal, sample_rate)``
|
47 |
+
to get a correctly converted signal in most cases.
|
48 |
+
|
49 |
+
Arguments
|
50 |
+
---------
|
51 |
+
wavs : torch.tensor
|
52 |
+
Batch of waveforms [batch, time, channels] or [batch, time]
|
53 |
+
depending on the model. Make sure the sample rate is fs=16000 Hz.
|
54 |
+
wav_lens : torch.tensor
|
55 |
+
Lengths of the waveforms relative to the longest one in the
|
56 |
+
batch, tensor of shape [batch]. The longest one should have
|
57 |
+
relative length 1.0 and others len(waveform) / max_length.
|
58 |
+
Used for ignoring padding.
|
59 |
+
normalize : bool
|
60 |
+
If True, it normalizes the embeddings with the statistics
|
61 |
+
contained in mean_var_norm_emb.
|
62 |
+
|
63 |
+
Returns
|
64 |
+
-------
|
65 |
+
torch.tensor
|
66 |
+
The encoded batch
|
67 |
+
"""
|
68 |
+
# Manage single waveforms in input
|
69 |
+
if len(wavs.shape) == 1:
|
70 |
+
wavs = wavs.unsqueeze(0)
|
71 |
+
|
72 |
+
# Assign full length if wav_lens is not assigned
|
73 |
+
if wav_lens is None:
|
74 |
+
wav_lens = torch.ones(wavs.shape[0], device=self.device)
|
75 |
+
|
76 |
+
# Storing waveform in the specified device
|
77 |
+
wavs, wav_lens = wavs.to(self.device), wav_lens.to(self.device)
|
78 |
+
wavs = wavs.float()
|
79 |
+
|
80 |
+
# Computing features and embeddings
|
81 |
+
outputs = self.mods.wav2vec2(wavs)
|
82 |
+
|
83 |
+
# last dim will be used for AdaptativeAVG pool
|
84 |
+
outputs = self.mods.avg_pool(outputs, wav_lens)
|
85 |
+
# # print(outputs.shape)
|
86 |
+
outputs = outputs.view(outputs.shape[0], -1)
|
87 |
+
# print(outputs.shape)
|
88 |
+
|
89 |
+
|
90 |
+
|
91 |
+
return outputs
|
92 |
+
|
93 |
+
def classify_batch(self, wavs, wav_lens=None):
|
94 |
+
"""Performs classification on the top of the encoded features.
|
95 |
+
|
96 |
+
It returns the posterior probabilities, the index and, if the label
|
97 |
+
encoder is specified it also the text label.
|
98 |
+
|
99 |
+
Arguments
|
100 |
+
---------
|
101 |
+
wavs : torch.tensor
|
102 |
+
Batch of waveforms [batch, time, channels] or [batch, time]
|
103 |
+
depending on the model. Make sure the sample rate is fs=16000 Hz.
|
104 |
+
wav_lens : torch.tensor
|
105 |
+
Lengths of the waveforms relative to the longest one in the
|
106 |
+
batch, tensor of shape [batch]. The longest one should have
|
107 |
+
relative length 1.0 and others len(waveform) / max_length.
|
108 |
+
Used for ignoring padding.
|
109 |
+
|
110 |
+
Returns
|
111 |
+
-------
|
112 |
+
out_prob
|
113 |
+
The log posterior probabilities of each class ([batch, N_class])
|
114 |
+
score:
|
115 |
+
It is the value of the log-posterior for the best class ([batch,])
|
116 |
+
index
|
117 |
+
The indexes of the best class ([batch,])
|
118 |
+
text_lab:
|
119 |
+
List with the text labels corresponding to the indexes.
|
120 |
+
(label encoder should be provided).
|
121 |
+
"""
|
122 |
+
outputs = self.encode_batch(wavs, wav_lens)
|
123 |
+
#outputs = self.CH(wavs, wav_lens)
|
124 |
+
outputs = self.mods.output_mlp(outputs)
|
125 |
+
out_prob = self.hparams.softmax(outputs)
|
126 |
+
score, index = torch.max(out_prob, dim=-1)
|
127 |
+
text_lab = self.hparams.label_encoder.decode_torch(index)
|
128 |
+
return out_prob, score, index, text_lab
|
129 |
+
|
130 |
+
def CH(self, wavs, wav_lens=None):
|
131 |
+
import torch
|
132 |
+
import torch.nn.functional as F
|
133 |
+
import soundfile as sf
|
134 |
+
from fairseq import checkpoint_utils
|
135 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
136 |
+
|
137 |
+
model_path = "D:\pycharm2020\code\yuyin_ChineseWav2vec\pretrained_models\Chinses_hubert\\chinese-hubert-large-fairseq-ckpt.pt"
|
138 |
+
wav_path = wavs
|
139 |
+
|
140 |
+
def postprocess(feats, normalize=False):
|
141 |
+
if feats.dim() == 2:
|
142 |
+
feats = feats.mean(-1)
|
143 |
+
|
144 |
+
assert feats.dim() == 1, feats.dim()
|
145 |
+
|
146 |
+
if normalize:
|
147 |
+
with torch.no_grad():
|
148 |
+
feats = F.layer_norm(feats, feats.shape)
|
149 |
+
return feats
|
150 |
+
|
151 |
+
print("loading model(s) from {}".format(model_path))
|
152 |
+
models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task(
|
153 |
+
[model_path],
|
154 |
+
suffix="",
|
155 |
+
)
|
156 |
+
print("loaded model(s) from {}".format(model_path))
|
157 |
+
print(f"normalize: {saved_cfg.task.normalize}")
|
158 |
+
|
159 |
+
model = models[0]
|
160 |
+
model = model.to(device)
|
161 |
+
model = model.half()
|
162 |
+
model.eval()
|
163 |
+
|
164 |
+
# wav, sr = sf.read(wav_path)
|
165 |
+
# feat = torch.from_numpy(wav_path).float()
|
166 |
+
feat = postprocess(wav_path, normalize=saved_cfg.task.normalize)
|
167 |
+
feats = feat.view(1, -1)
|
168 |
+
padding_mask = (
|
169 |
+
torch.BoolTensor(feats.shape).fill_(False)
|
170 |
+
)
|
171 |
+
inputs = {
|
172 |
+
"source": feats.half().to(device),
|
173 |
+
"padding_mask": padding_mask.to(device),
|
174 |
+
}
|
175 |
+
|
176 |
+
with torch.no_grad():
|
177 |
+
logits = model.extract_features(**inputs)
|
178 |
+
outputs = self.mods.avg_pool(logits[0], wav_lens)
|
179 |
+
# # print(outputs.shape)
|
180 |
+
outputs = outputs.view(outputs.shape[0], -1)
|
181 |
+
# print(outputs.shape)
|
182 |
+
|
183 |
+
return outputs
|
184 |
+
|
185 |
+
|
186 |
+
|
187 |
+
def classify_file(self, path):
|
188 |
+
"""Classifies the given audiofile into the given set of labels.
|
189 |
+
|
190 |
+
Arguments
|
191 |
+
---------
|
192 |
+
path : str
|
193 |
+
Path to audio file to classify.
|
194 |
+
|
195 |
+
Returns
|
196 |
+
-------
|
197 |
+
out_prob
|
198 |
+
The log posterior probabilities of each class ([batch, N_class])
|
199 |
+
score:
|
200 |
+
It is the value of the log-posterior for the best class ([batch,])
|
201 |
+
index
|
202 |
+
The indexes of the best class ([batch,])
|
203 |
+
text_lab:
|
204 |
+
List with the text labels corresponding to the indexes.
|
205 |
+
(label encoder should be provided).
|
206 |
+
"""
|
207 |
+
waveform = self.load_audio(path)
|
208 |
+
# Fake a batch:
|
209 |
+
batch = waveform.unsqueeze(0)
|
210 |
+
rel_length = torch.tensor([1.0])
|
211 |
+
outputs = self.encode_batch(batch, rel_length)
|
212 |
+
outputs = self.mods.output_mlp(outputs).squeeze(1)
|
213 |
+
out_prob = self.hparams.softmax(outputs)
|
214 |
+
score, index = torch.max(out_prob, dim=-1)
|
215 |
+
text_lab = self.hparams.label_encoder.decode_torch(index)
|
216 |
+
return out_prob, score, index, text_lab
|
217 |
+
|
218 |
+
def forward(self, wavs, wav_lens=None, normalize=False):
|
219 |
+
return self.encode_batch(
|
220 |
+
wavs=wavs, wav_lens=wav_lens, normalize=normalize
|
221 |
+
)
|
pretrained_path/label_encoder.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b7bb126752aed05089d4f211eaf82ceab98c4a17c2813774b58091441064c6ad
|
3 |
+
size 73
|
pretrained_path/label_encoder.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
'depress' => 0
|
2 |
+
'non_depress' => 1
|
3 |
+
================
|
4 |
+
'starting_index' => 0
|
pretrained_path/model.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:293e54327bc8c42df3028fa045a7c043a5c25d05085ae32d5af1c022820be1ae
|
3 |
+
size 8997
|
pretrained_path/wav2vec2.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:31fd83ae093ab6a97f04cd8612d419cca8fc78a0cf9f4b11b35dfc29dd5cd5e3
|
3 |
+
size 1261924189
|
speech.py
CHANGED
@@ -75,4 +75,4 @@ demo = gr.Interface(
|
|
75 |
],
|
76 |
)
|
77 |
|
78 |
-
demo.launch()
|
|
|
75 |
],
|
76 |
)
|
77 |
|
78 |
+
demo.launch(share=True)
|